EmoRoBERTa Open-Source Sentiment Classification Model - Precise Identification of 28 Sentiment Categories

Emoroberta

Developed by arpanghoshal

A fine-grained sentiment classification model based on the RoBERTa architecture, trained on the GoEmotions dataset, capable of recognizing 28 emotion categories.

Text Classification

Transformers

EnglishOpen Source License:MIT #Fine-grained Sentiment Analysis #Reddit Comment Classification #RoBERTa Optimization

Downloads 21.47k

Release Time : 3/2/2022

Model Overview

This model is specifically designed for text sentiment analysis, capable of identifying 28 different emotion categories including happiness, anger, sadness, etc. It is suitable for emotion recognition in scenarios such as social media comments and user feedback.

Model Features

Fine-grained Emotion Recognition

Supports fine-grained classification of 28 emotion categories, far exceeding traditional three-category models (positive/negative/neutral).

Improved Based on RoBERTa

Utilizes an optimized RoBERTa architecture, enhancing performance through hyperparameter tuning and expanded training data.

Optimized for Reddit Comments

Trained on the GoEmotions dataset (58,000 Reddit comments), making it particularly suitable for social media text analysis.

Model Capabilities

Text Sentiment Classification

Fine-grained Emotion Recognition

Social Media Comment Analysis

Use Cases

Social Media Analysis

User Comment Sentiment Analysis

Analyze the emotional tendencies of user comments on platforms like Reddit

Can identify 28 specific emotional states

Customer Feedback Analysis

Product Feedback Emotion Classification

Perform fine-grained sentiment classification on customer feedback

Accurately identifies details of customer satisfaction (e.g., gratitude, disappointment, etc.)

🚀 EmoRoBERTa: Text Classification for Emotion Analysis

This project uses RoBERTa to perform text classification on the GoEmotions dataset, aiming to identify 28 different emotions in text.

🚀 Quick Start

The following is a basic example of using the EmoRoBERTa model for emotion analysis:

from transformers import RobertaTokenizerFast, TFRobertaForSequenceClassification, pipeline

tokenizer = RobertaTokenizerFast.from_pretrained("arpanghoshal/EmoRoBERTa")
model = TFRobertaForSequenceClassification.from_pretrained("arpanghoshal/EmoRoBERTa")

emotion = pipeline('sentiment-analysis', 
                    model='arpanghoshal/EmoRoBERTa')

emotion_labels = emotion("Thanks for using it.")
print(emotion_labels)

Output

[{'label': 'gratitude', 'score': 0.9964383244514465}]

✨ Features

Rich Emotion Labels: The model can classify text into 28 different emotions, including admiration, amusement, anger, etc., as well as a neutral category.
Powered by RoBERTa: Leveraging the advanced RoBERTa architecture, which has better generalization ability compared to BERT.

📚 Documentation

What is GoEmotions

The GoEmotions dataset labels 58000 Reddit comments with 28 emotions:

admiration, amusement, anger, annoyance, approval, caring, confusion, curiosity, desire, disappointment, disapproval, disgust, embarrassment, excitement, fear, gratitude, grief, joy, love, nervousness, optimism, pride, realization, relief, remorse, sadness, surprise + neutral

What is RoBERTa

RoBERTa builds on BERT’s language masking strategy and modifies key hyperparameters in BERT. It removes BERT’s next - sentence pretraining objective and trains with much larger mini - batches and learning rates. RoBERTa was also trained on an order of magnitude more data than BERT for a longer time, allowing its representations to generalize better to downstream tasks compared to BERT.

🔧 Technical Details

Hyperparameters

Property	Details
Learning rate	5e-5
Epochs	10
Max Seq Length	50
Batch size	16
Warmup Proportion	0.1
Epsilon	1e-8

Results

The best result of Macro F1 is 49.30%.

📄 License

This project is licensed under the MIT license.

linkedin.com/in/arpanghoshal

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご